No, I am not going to tell you to use throw; instead of throw e; I am going to talk about exception messages, assumptions, and pain.
Exception hierarchies are useful in many ways, mostly because they bring order to the way we handle exceptions. We can catch a specific exception, or a root exception in a hierarchy, and hanlde them specifically. But, one of the usages of exception hierarchies is to add additional data to an exception. In many cases, this is very useful data, such as the SQL error code or the details node of in a soap fault or the list of assemblies that could not be loaded.
Do you know what this three datums has in common?
- They are very useful
- They do not show in ex.ToString()
Guess what is going to be shown in any log, error message, etc?
You got that right, the ex.ToString() output!
If you have additional information in the exception, it must be available on the exception afterward. Trying to diagnose assembly load failures is driving me mad.
Imagine finding things like this in the log:
- "ReflectionTypeLoadException: Unable to load one or more of the requested types. Retrieve the LoaderExceptions property for more information."
- "SoapException: Server was unable to process request"
Imagine gettting one of those during dev, you have no way of knowing where this is happening. Often you can't even set a breakpoint in the code there, and inspect the exception, because it is handled inside some library code. ASP.Net is a good example of how this can happen, and where this is a highly annoying issue to work with.
To summrise, if you create exceptions, make sure to remember a simple rule, everything should go in the ex.ToString().